基于建筑物平面轮廓特征的城市居住建筑识别模型  

An Urban Residential Buildings Recognition Model based on Building Footprint Geometry Feature Identification

作　　者：邹志翀 张清荷[1,2] 翟凤勇 Zou Zhichong;Zhang Qinghe;Zhai Fengyong

机构地区：[1]哈尔滨工业大学建筑学院 [2]哈尔滨工业大学寒地城乡人居环境科学与技术工业和信息化部重点实验室 [3]哈尔滨工业大学管理学院

出　　处：《建筑与文化》2021年第7期225-228,共4页Architecture & Culture

基　　金：国家重点研发计划项目(项目编号:2018YFC0704705)。

摘　　要：建筑物的功能信息有助于深入理解城市的运作方式,在智慧城市构建、城市规划、灾害管理、建筑能耗模拟等领域起到重要支撑作用。文章构建了一套基于形态特征认知的模型,通过开源数据实现大规模的城市居住建筑识别。该模型在以建筑物的平面轮廓特征作为训练数据的基础上,通过对居住建筑典型形态类型的定义和样本提取、训练,构建建筑物功能识别模型。通过在两个城市新区5880栋建筑对算法有效性的验证分析,本模型的识别准确率分别为82.2%和74.1%。The functional information of urban buildings helps to have in a profound understanding of rules in a city. It plays a key role in fields such as smart city construction, urban planning, disaster emergency management and building energy consumption simulation. Based on morphological characteristics, this paper constructs a method to identify urban residential buildings in large scale by using open source map data. Based on extracting geometry features as training data from the raw building footprints, this proposed method is constructed by two procedures of defining typical morphological types of residential buildings, and sample extraction. Taking two new urban areas with total 5880 buildings as verification and validation, the recognition accuracy of this proposed model is 82.2% and 74.1%.

关 键 词：城市居住建筑 建筑足迹数据 随机森林模型 监督分类 建筑形态类型学 

分 类 号：TU241[建筑科学—建筑设计及理论]